Analysis of the microarray gene expression for breast cancer progression after the application modified logistic...

Area of Science:

• Bioinformatics
• Genomics
• Cancer Research

Background:

• Microarray technology globally analyzes gene expression, but faces challenges like low sample size and high dimensionality.
• Traditional methods often reduce data complexity, potentially losing valuable information.
• Classifying breast cancer subtypes from microarray data remains a critical challenge.

Purpose of the Study:

• To introduce a novel logistic regression-based model for classifying breast cancer tumor samples using comprehensive microarray gene expression data.
• To avoid feature selection or data matrix reduction, preserving all available gene expression information.
• To identify potential novel biomarkers and therapeutic targets for breast cancer.

Main Methods:

• Developed a logistic regression model that incorporates all gene expression features without prior data reduction.
• Applied the model to classify breast cancer sample datasets from Gene Expression Omnibus (GEO) series GSE65194, GSE20711, and GSE25055.
• Explored all data combinations, including various breast cancer subtypes, to assess model performance.

Main Results:

• Achieved a minimum classification performance of 80% (sensitivity and specificity) across diverse breast cancer datasets.
• Identified novel genes with extreme parameter values, suggesting their potential as breast cancer prediction genes.
• Highlighted genes with tumor suppressor profiles (e.g., PKN2, MKL1) and oncogenic profiles (e.g., MTR, ITGA2B).

Conclusions:

• The proposed logistic regression model effectively classifies breast cancer subtypes without data reduction, offering high accuracy.
• The methodology reveals previously unstudied genes and confirms known biomarkers, providing insights into breast cancer.
• Identified potential breast cancer prediction genes and therapeutic targets warranting further clinical investigation.